Options for wiring ground

[Joel Pitt]

Kiwiburn is approaching and I'm working on a new project:

https://github.com/ferrouswheel/tensegrity

(A follow on from the previous year's project:
https://github.com/ferrouswheel/sensorium )

My question is related to wiring up a common ground. The default
wiring diagram here:

https://raw.githubusercontent.com/scanlime/fadecandy/master/doc/images/system-diagram-1.png

... and it shows a single strip, with the ground connected via the led strip.

The Sensorium had the ground coming from a PSU to a choc block and
wired in parallel to each strip. The PSU also had ground going back to
the control box with two teensy 3.0s. There was only the data line
going from control box to each led strip.

Because the LED strips are not adjacent to the fadecandies I'm using,
I'm wondering whether I can save some wiring by connecting a ground to
the fc and then soldering all the fc grounds together? All strips will
be powered from PSUs with a common ground.

Ok, so that's the main question, and my rudimentary understanding says
this will probably work (correct me if I'm mistaken please!). However,
I'm more interested in the caveats of doing it this way. e.g. would
running a ground with the data signal directly from fc to led strip
result in lower noise?

The lengths of wiring for data I'm using are just over 1m at the maximum.

Any hints/comments appreciated :-)

Joel

[Micah Elizabeth Scott]

The shortest answer I can give is "probably fine", but here's a much longer answer starting with a picture. Maybe it helps in planning your wiring.

This is a really simplified version of the Fadecandy and WS2812 to illustrate the two loops that electrical current flow around, labeled (1) and (2).

The Fadecandy and WS2812 each do a lot of things, but for this problem we can simplify everything down to basic components like switches and power supplies. Don't worry if the schematic symbols are unfamiliar; I've tried to label the important parts.

Loop (1) goes through the Fadecandy itself (simplified here as a power supply and a switch) and through the Data input on your first LED, which acts roughly like a resistor connected to the WS2812's ground.

Loop (2) goes through your large +5V supply, into a smart on-off switch inside the WS2812, into the actual LED (red, green, or blue work the same way), and back through the strip's ground.

If all of your wires were ideal superconductors it wouldn't really matter where the grounds are wired or what gauge cables you use.  Since these are real wires, they aren't 100% efficient. When current flows, some voltage is lost, converted to heat and a tiny bit of electromagnetic radiation.

By Ohm's Law, the losses (in volts) are proportional to the current (in amps) and the resistance of the wire (in Ohms). This is the same formula you'd use to calculate losses given a particular load and wire gauge.

That switch on the LED complicates matters. Because the LEDs are all turning on and off really fast, the current in loop (2) is changing rapidly. The capacitors on your LED strip will smooth out these changes a little bit, but it's still going to be fairly tumultuous in there. The rapidly changing current translates (Ohm's law again) into a rapidly changing voltage. Other factors like inductance can magnify this voltage even more.

On its own, this voltage fluctuation wouldn't be a big deal. It would cause your LEDs to change brightness only very very slightly. We'll have problems though if this disrupts the data signal flowing through loop (1).

The WS2811's data input senses the voltage difference as a very small current flows from the data pin to the ground pin. If loops (1) and (2) share wire, the voltage that develops across that wire due to the large current in loop (2) will be added/subtracted to the voltage the input "sees".

And this might be fine. It's only a problem if all of the total interference is enough to flip a 0 to a 1 or vice-versa.

The setup you're describing is similar to what I did on Ecstatic Epiphany. In that project I ran three wires from each strip back to a junction area along one side. Each of those junctions had its own ground ("star" topology out to all the strips), and the Fadecandy ground was connected to that junction with one small wire.

The length that matters most is the section of wire where both of the current loops are sharing one conductor. In Ecstatic, that was about 2 meters at most. As far as I can tell, that hasn't been causing any problems.

Cheers!

–m

Joel

--
You received this message because you are subscribed to the Google Groups "Fadecandy" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fadecandy+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Joel Pitt]

Thank you hugely for taking the time to explain this to me!

While I'm appreciative of many great resources and guides out there that explain "how" to do various hobbiest electronics, I also love understanding "why".

So thanks for this and thanks for fadecandy :-)